137091 research outputs found
Sort by
Autonomous Landing of a Multirotor on a Mobile Vehicle Using Infrared Beacons
No full textThis dissertation contains a multi-faceted study into the problem of autonomously landing a multirotor uncrewed aerial vehicle (UAV) onto a moving ground vehicle under challenging environmental conditions and functionally limiting constraints. First, the dissertation will introduce the problem and related work from the literature. Next, the various components of a model of the mobile landing problem will be developed and described. Results from a simulation study performed with these models are then presented. The final main sections of the dissertation cover the design, fabrication, and test results from two generations of mobile landing systems. These results constitute an addition to the state-of-the-art literature on mobile landings because of the speed of landings achieved in a relatively unstructured outdoor environment. To conclude, these contributions to the academic literature are summarized, and future research directions are laid out.
Many other solutions to autonomously landing on a moving ground vehicle have been developed in the literature, some of them to great functional effect. However, most of these solutions rely on global position systems (GPS), communication between vehicles, or visual targets requiring precise camera calibration based on current environmental conditions. These requirements contrast the reality that many of the most promising applications for teams of UAVs and ground vehicles lie in unstructured environments. In search and rescue and military contexts, communications, GPS, and lighting conditions are far from guaranteed. Therefore, the UAVs documented in this dissertation do not require GPS or communications with the ground vehicle; instead, their relative localization systems rely on infrared beacons. This allows robust landing operations, demonstrated on off-road surfaces and in any lighting condition, including complete darkness.Ph.D.Mechanical Engineerin
Sustainability and Resiliency in Airport Energy Infrastructure: A Multidisciplinary Methodology for Optimizing Building Operations
No full textPresented at the 2025 AIAA SCITECH forum.There is a striking di!erence in the domain of environmental consciousness: a large amount of greenhouse gases is produced by commercial buildings compared to what is commonly perceived in the transportation sector. The International Panel on Climate Change has labeled buildings as the third largest contributor of greenhouse emissions with at least 17.5%, whereas the aviation industry merely contributes 2%. Hence, there is an urgent need to decarbonize buildings to enhance sustainability, a sound path towards mitigating e!ects of climate change. Moreover, continued power supply to critical buildings such as hospitals and airports, is important to ensure life safety of their users. Power outages in such buildings disrupt operations resulting to substantial financial losses. For instance, the 2017 outage at the Hartsfield Jackson Atlanta Airport in the United States demonstrated the result — a 11-hour period without power which brought Delta airlines about $50M estimated in loss. From a system-of-systems perspective, this paper investigates decarbonizing of buildings and infrastructure by considering aspects of sustainability, resiliency, and a!ordability. A detailed account of scenarios involving power outages, building HVAC systems and the demand of electric vehicles is balanced against the amount of power from the main grid and distributed energy resources including photovoltaics,storage systems and power generators. The aim of this study is to provide a financially optimum combination of grid energy and DERs (Distributed Energy Resources). As such, this study utilizes a proposed methodology that employs use of multi-variate regression models to integrate a building HVAC system modeled using Simcenter Amesim in a representative thermal envelope with an optimization tool developed by the US National Renewable Energy Laboratory. These modeling tools are used to create a parametric and interactive tool that assists stakeholders in assessing tradeo!s for building energy sourcing to meet power demands even during power outages and assess its impacts financially and environmentally
Biomechanics and Risk of Coronary Obstruction in Transcatheter Aortic Valve Replacement
No full textTAVR has rapidly evolved into the preferred method of aortic valve replacement, taking over SAVR in all age groups due to expansion of indication to all patients with aortic stenosis and great outcomes in long term studies. However, several complications have been reported that may occur as a result of TAVR and therefore careful procedural planning becomes essential to ensure good patient outcomes. Computational modeling can provide accurate visualizations of the post-TAVR configurations of the device and the native anatomy to extract information about the potential for complications. However, computational modeling of TAVR is not regularly used in clinical practice and one of the reasons could be the lack of data on validation of the computational models in predicting the device deformation in a range of clinical anatomies. To treat patients where the only treatment option is TAVR, surgeons are adopting strategies that help with mitigating the risk of complications such as changing the deployment depth, balloon volume or post-dilatation of the bioprosthesis and laceration of the native or bioprosthetic valve leaflets. However, it is not fully understood the impact of such adaptations on the device itself or the potential for other complications. Coronary obstruction is a serious procedural complication associated with high mortality rates. Proper standardized assessment of the risk of coronary obstruction during procedural planning is necessary and current clinical guidelines fail to achieve sufficient accuracy in predicting the complication. The studies contained in this thesis document aim to resolve these clinical questions and deficiencies with an overarching goal of adding to the knowledge of biomechanics of TAVR in native and bioprosthetic aortic valves.
In the first aim, development and clinical validation of deployment models of TAVR valves currently in use clinically in patients with failed native tricuspid aortic valves is described. Simulations of the THV deployment using finite element methods showed excellent agreement with the post-TAVR CT images. Adjustments to the deployment methods were tested. Changes to the implantation depth had no impact on the THV expansion or shape in native aortic valves. Overfilling of the deployment balloon in SAPIEN showed improvement in device expansion and underfilling showed an increase in potential for leaflet thrombosis. Post-dilatation of Evolut increased expansion in the functional region. In the second aim, the developed models were applied to predict deformation in valve-in-valve TAVR. Lower deployment worsened the functional area in Evolut, and a higher implant decreased inflow area for the SAPIEN. Both devices were found to expand better with laceration of bioprosthetic aortic valve leaflets. In the third aim, coronary obstruction predictive models based on THV deployment simulations were developed and validated against post-TAVR outcomes in both native and valve-in-valve TAVR. The impact of transcatheter valve type on the risk of coronary obstruction was studied. The outcomes of this thesis can help clinicians better visualize transcatheter valve deformations in native and valve-in-valve TAVR, better understand the impact of procedural adaptions and optimize the selection of THV to minimize the risk of coronary obstruction in every patient.Ph.D.Bioengineerin
What Happens When a Robot Lies to You? Investigating Aspects of Prosocial Intelligent Agent Deception Towards Humans
No full textPeople across many societies are explicitly taught some form of the adage “honesty is the best policy”, but is that a lie? Telling the truth is not always helpful, and lying is not always harmful. In truth, everyone lies. We lie to help ourselves, and we lie to help others. We lie in both serious and inconsequential situations. Lying is a foundational part of how people interact with each other, and accepted members of society are successfully able to navigate the highly nuanced norms of social deception.
Robots and artificially intelligent (AI) systems are increasingly being placed within our societies, and in some contexts, they are expected to interact with humans socially. People must trust that robots are functionally competent to complete tasks while also being socially competent to understand social conventions that may favor particular strategies over others. If people often successfully choose lying to be the best policy in certain situations, it then follows that a robot, that is designed to learn from humans and exhibit social competency, may replicate expected lying behavior as it becomes fully integrated into social settings.
In this thesis I explore robots that lie to benefit others and how deception influences people’s interactions and perceptions of robots. My work examines how managing expectations, the influence of agent design and presence, and the aftermath of deception shape human responses, while also exploring how people interact with autonomous deceptive agents over time.Ph.D.Computer Scienc
Synthetic Medical Micro/Nanorobots for In Vivo Biomedical Applications
No full textPresented on September 3, 2025 at 12:15 p.m. in the Marcus Nanotechnology Building, Room 1116.Wei Gao is a Professor of Medical Engineering at the California Institute of Technology. Professor Gao's primary research interest is in the development of novel bioelectronic devices for personalized and precision medicine: wearable and flexible biosensors that can analyze the various biomarkers in body fluids for real-time continuous health monitoring and early diagnosis, and synthetic micro/nanomachines for rapid drug delivery and precision surgery. His research thrusts include fundamental materials innovation as well as practical device and system level applications in translational medicine.Runtime: 63:03 minutesThe 1966 film Fantastic Voyage inspired the concept of miniature machines traversing the human body to diagnose and treat disease. This once-futuristic vision is now being realized through the development of synthetic micro/nanorobots engineered for in vivo biomedical applications. In this talk, I will highlight recent advances in the design of microrobots powered by bioavailable fuels or externally applied fields such as ultrasound and magnetism. These systems enable a range of sophisticated capabilities, including biosensing, targeted drug delivery, bioimaging, and cellular isolation. I will discuss our recent work on ultrasound-propelled microrobots that achieve deep-tissue navigation and controllable drug release guided by real-time ultrasound imaging. In parallel, I will present biofuel-powered micro/nanomotors designed for effective tissue penetration and programmable therapeutic delivery, particularly in oncology settings. These platforms integrate responsive materials, wireless control, and clinical imaging interfaces, transforming synthetic motors into intelligent, adaptable therapeutic agents. Our results reveal the vast potential of medical microrobots in addressing unmet clinical needs across gastrointestinal disorders, inflammation, cancer, and precision medicine
Metabolomics of Inflammatory Bowel Disease in African American Patients
No full textInflammatory Bowel Disease (IBD) in African American (AA) and individuals of European Ancestry (EA) is steadily growing. Unfortunately, IBD studies underrepresent AA, compromising our understanding of disease etiology and progression in the subpopulation. IBD affects 96 of 100,000 AA every year, so generation of numerous well-balanced and matched biospecimens for metabolomic studies is challenging. Metabolomics provides the most immediate snapshot of the IBD phenotype, showing contrasting pathogenesis in AA vs. EA individuals. We here present a new Dynamic SQUAD approach where LC-MS experiments are performed in a sequential fashion as more biospecimens become available. This approach builds a targeted metabolite panel that reflects IBD pathway alterations from the non-targeted data block produced by the previous SQUAD LC-MS batch. We conducted SQUAD LC-MS experiments on biopsies from IBD patients to measure targeted and non-targeted data for different analytes using a Thermo ID-X tribrid mass spectrometer. A literature metanalysis indicated specific metabolic reactions that can differentiate Crohn’s disease (CD) from healthy patients. Pathways altered included amino acid metabolism, tryptophan metabolism, microbial metabolism, and energy metabolism, indicating dysbiosis of the gut microbiome, showing a decrease in the diversity of microbes/metabolites in the gut1. Additionally, metabolic modeling indicated that inflamed CD patient ileal tissue displays a distinct metabolic signature compared with non-inflamed tissue2. Subsequently, we sought to experimentally confirm whether the CD ileum contained metabolic changes versus controls using non-targeted LC-MS of a small external pilot cohort, complementary to in silico metabolic modeling to characterize whether the changes in metabolic pathways correlate to specific changes in specific lipids and/or metabolite composition. The top discriminant analytes were also added to the first targeted SQUAD panel. These ongoing studies will help us probe metabolomics alterations in the mucosal tissue to further our insights on IBD differences across populations.M.S.Chemistry and Biochemistr
Image Guided High Precision Robotic Positioning in MRI for Medical Applications
No full textMagnetic Resonance Imaging (MRI) is a powerful diagnostic tool that offers advanced visualization of human tissue, increasingly used to guide medical procedures such as biopsies and interventions. Nevertheless, navigation in the MRI environment remains challenging due to material, actuator, and sensor restrictions as well as scan time and cost of use.
This work presents methods for ensuring high precision robotic positioning in MRI for use in emerging applications through three distinct aims. In the first aim, an MRI-analogous test bench implementing Position Sensitive Devices (PSDs) is established to measure the positioning performance of a previously developed MRI compatible robot, circumventing limitations of MRI resolution and scan time, validating the capability of MRI guided robot navigation methods. In the second aim, the validated high-precision navigation method is leveraged to enable the application of multi-image Super Resolution (SR) algorithms to construct enhanced resolution in-plane MRI slices, leading to improved positioning precision exceeding the limits of the native MRI resolution. In the third aim, a data-driven gain estimation control method is established to compensate for resistive forces and improve open-loop positioning accuracy when the robot end-effector navigates through a complex fluid medium. A novel acousto-optic sensor is integrated into the system to measure impacts of radio-frequency waves on temperature and e-field distribution around medical implants in MRI. Improved open-loop control reduces the number of MRI scans needed for high accuracy positioning, reducing experiment and procedure time, allowing for navigation to larger number of points and expanded data collection within a set time frame. These developments enable the assessment of medical implant safety in MRI through high accuracy positioning needed to properly understand dissipation of temperature and e-field around conductive structures.Ph.D.Robotic
AutoCurate : Automating Domain-Specific Dataset Curation for Large Language Models
No full textLarge Language Models (LLMs) have demonstrated remarkable performance on open-domain tasks, yet they often struggle with factual accuracy and terminology in specialized domains such as finance, law, or medicine. Existing approaches to building domain-specific LLMs typically rely on fully curated in-domain corpora, which are expensive and labor-intensive to assemble at scale. In this work, we propose a novel, scalable pipeline for domain-specific dataset curation that minimizes manual intervention. Our method combines topic modeling and BM25-based filtering to iteratively expand a small seed corpus into a high-quality, domain-relevant dataset drawn from large-scale generic corpora. We demonstrate the effectiveness of our approach by curating a financial dataset exceeding 100 billion tokens from the Dolma corpus. We will publicly release the datasets and an industrial-grade implementation of our pipeline to facilitate its application across other domains. Overall, our work presents a practical and extensible solution for building high-quality domain-specific training corpora, advancing the development of reliable, specialized LLMs.UndergraduateComputer Scienc
Increasing Resilience of Intermodal Freight Transport Networks–Key Challenges in Disruption Handling and Requirements for Digital Solutions
No full textThe Eurozone faces growing economic and environmental challenges, with supply chain disruptions causing losses of over EUR 112.7 billion in 2021. Climate change, geopolitical risks, regulatory shifts, and infrastructure weaknesses strain intermodal freight transport, highlighting the need for digital solutions to enhance resilience and efficiency.
This paper examines key challenges in intermodal freight transport, including disruption triggers, network vulnerabilities, and inefficiencies in disruption management. Extreme weather, and capacity shortages impact both performance and sustainability. Using the Total Quality Framework (TQF), the research includes interviews with 23 stakeholders from 10 countries, focus workshops, and surveys. The analysis reveals shortcomings in real-time data integration, interoperability, and disruption response. Regulatory fragmentation and low digital maturity hinder resilience strategies. Addressing these gaps requires harmonized data frameworks, improved interoperability, and the use of collaborative digital platforms.
The Horizon Europe project ReMuNet leverages intelligent algorithms and digital platforms to enhance multimodal networks, optimize route planning, and improve disruption response, contributing to the vision of the Physical Internet
Layout-Agnostic Change Point Detection for Enhanced Human Activity Recognition in Smart Homes
No full textHuman Activity Recognition (HAR) in smart home environments relies heavily on accurate temporal segmentation, yet most existing methods use fixed, overlapping windows that often blur activity boundaries and degrade downstream classifier performance. This thesis introduces two layout-agnostic change point detection (CPD) frameworks, TAD and TCPC, that achieve state-of-the-art results across multiple smart home datasets. TAD (Textual Descriptions of Sensor Triggers-based Activity Distribution) Change Point Detection uses semantically rich sensor metadata and a BiLSTM HAR classifier to detect changes in activity distributions between non-overlapping windows, while TCPC (Textual Descriptions of Sensor Triggers-based Contrastive Predictive Coding) Change Point Detection eliminates the need for labeled data by using a contrastive model that identifies change points through variations in learned latent representations. Evaluated on four CASAS smart homes (Aruba, Milan, Kyoto, and Cairo), both methods consistently outperformed statistical and embedding-based baselines, improving the geometric mean of true positive and true negative rates for change point detection by 1.20% to 16.45%, and enhancing downstream classification with a BiLSTM model by 1.66% to 8.18% in Macro F1 score. Leveraging lightweight sentence-transformer embeddings and modest recurrent architectures, these approaches are suitable for real-time deployment and offer a practical, representation-driven alternative to traditional segmentation techniques, paving the way for future work involving transformers, richer datasets, and energy-efficient models.UndergraduateComputer Scienc